s02 e08 Absolute Randomness

Welcome to the Big Cheese AI podcast. I'll be your host today. I'm Sean Hise. I'm here as always with my counterparts, Brandon Corbin. Good day, sir. And Jacob Wise. Hello. It's just us three today. We had a cancellation with our guests, so we're just hanging out. Yeah. You know, we've actually got a lot of cool things to talk about today. We're going to talk a little bit about some really interesting news, maybe not so interesting after further research. We've got we've got chat, cheap voice. we've got iOS 18.1 with the first Apple intelligence features now being released to anybody who likes to install betas, which I'm still on the fence about. We've got a new llama model. We've got search GPT. But the biggest news of the week is, and I know when reacted to it on the Slack channel, Taco Bell is rolling out AI in the drive-through. Yeah, I didn't read that article. Did they say which software were they going with Google? So check this out. So, well, no, I don't know. I don't know, but what I thought was interesting is that they are going to be able to be more accurate than, they're finding that just having AI listen to the orders, categorize it and map it to items, is saving the people in the stores a bunch of time and it's more accurate than a human doing it. So I thought like, is maybe like last month that they had done some trials, but then they turned it off. Yeah, I think that, I think they're actually gonna start rolling it out.

• Because there was something that happened.
• Other chains dropped it because, and they were using, there was a couple of big startups in that space. And there was another chain that we talked about two shows ago, and I forget now I'm blanking on it, but they were just like, it was causing a lot of confusion. And then like, there was like a TikTok. Yeah, like a TikTok person. No, I don't want to have millions.
• So when I go ask them, "Chile, Chile, easy, busy," the human always knows, I mean, Chile, she's burrito.
• Yeah. (laughing)
• They do. They, I will probably, they'll have to make it a best hallucination. Who doesn't love a surprise in their Taco Bell bag?
• Right.
• I miss Taco Bell.
• Yeah.
• Well, Taco Bell's so great because they have like five ingredients and they're just like hit the remix button. (laughing)
• Exactly.
• Some combination of these things.
• Well, it does kind of point to the fact that the, we continue to see the applications of just large language models really being where the, where the innovation is. and taking mental load off people, being the where it adds the most value. And if it adds value, trying not to deal with putting orders in the queue and dealing with customer service, it's a, that's something that actually can make people's lives easier. I don't think that's, I think that makes them more productive. I think it makes them, the lines go faster. It doesn't kill, I don't think that necessarily kills jobs or anything. I mean, it's just.
• Well, wasn't it too though, that like McDonald's was exploring about having fully remote, like when you were actually talking to somebody, it was somebody in a call center. Wendy's did it for sure. Yeah, yeah, yeah. I've felt that before, like the person at the line definitely wasn't there. Right, yeah. So they've obviously had this business problem in the past and-- I don't know, man. I think fast food restaurants are going in 10 years. I don't think they have any employees. Oh. Like maybe one or two, that's just like the manager who just goes in, make sure that everything's stocked, but the rest of it's just fully automated. 'Cause it's gonna get too expansive. Like if they ever actually increase the minimum wage from the $7 or whatever it is, you know, I mean, it's gonna become cost prohibitive.
• That's an idiot. - That's an idiot.
• You just go up to the ATM and you have a big aspirator. (laughing)
• Hulk Hogan on that Trump, on the Trump, whatever that thing he was at, literally was a scene straight out of idiosyncrasies.
• Dude, it was bananas.
• So last night we all went to, and our local, for those who are more of our national listeners, we were from Indianapolis, and Indianapolis has a lot of startups and ecosystems really built around marketing, but there's also a pretty good developer community. They had a shout out to Zach Cardwell, Andy Hackers through an event last night. We were hanging out with almost what, 250 developers last night?
• That was crazy.
• I think those 250 signed up. I bet 150 showed up, maybe, maybe more. I felt like more, yeah.
• What was really, really cool experience. I mean, obviously the industry's live and well. We talked to people that are gainfully employed. There wasn't a ton of AI conversation, but we definitely saw the antithesis of that, which one company I talked to was marketing their product because it doesn't use AI.
• Yeah.
• Because a lot of the, we've been hearing this, a lot of the boards are really concerned about security and privacy.
• Yeah.
• You're talking about your data. I thought that was interesting. You guys have any other takeaways?
• No, I was kind of surprised that we didn't have more, oh, and I did that thing too on LinkedIn where I asked, like, hey, are you an Indianapolis startup? That's doing AI, and we had a couple that reached out, and so if you're actually watching the show to talk about, we're gonna be saving that for when our other guest comes back in, we'll drop your names on it. Matt, you can come on anytime now. So, but I was surprised that there wasn't, there wasn't as many just like pure AI plays. It was also surprised that there was just not as many that are just even leveraging AI for different parts of their technology stack, right? Like, 'cause again, there's, I mean, you don't have to be an AI startup, you can just be a regular startup, but you can start trying to leverage AI to kind of improve things. But yeah, there just wasn't as much conversation about that as I would have expected.
• No, I didn't hear as much.
• I will say when I go to like the more junior developer meetups, it's all we talk about. (laughing) That junior Dather just like, what do I need to know?
• Right.
• And I feel like more seasoned people, like I use it, we all use it in our tool chain, like, you know, while we develop, but it's still just a piece of the puzzle. Yeah. And it used to be real. I mean, it's exciting, but it's just another tool. It's a calculator at this point. Yeah. So like quick, and I have young children, you know, been in and out of that networking game for years, but like my, one of the takeaways I had, and this is just more of an ecosystem thing is like we hit the ground hard around 2016 to 2017, and there was such a, there It was such a, it wasn't AI, but for us it was front-end tech. It was the dawn of React and view and all these different technologies that led people to build really cool, really, really impressive apps. And there was a lot of job growth out of that, a lot of startups getting created. I did sense a little bit of that theme kind of calming down a little bit. There's a lot of people that are out there kind of grinding, but do you think that there's a sense that the engineer or the developer is becoming less of this unbelievable opportunity and more of just another desk job kind of concept? Do you do? I've got that feeling. I'll try, man. I feel like, so we've been running an agency and you've been developing for a long time and I feel like there was a golden age of become a developer. The dream was clear. You don't love your job. You're not super passionate about it. You're not making a lot of money. If you go to a boot camp or self-teach for a few months and learn a solid front end or any coding application or technology, you can get a great job. It's high paying. It's pretty easy, quote unquote to do. That is no longer the case.
• No, it's not.
• And I think that is translating not just to the junior or kind of entry level positions, but sort of higher up the chain where every engineer I talk to is more of a hybrid position at this point, where I actually did have a conversation last night with someone about what do you think about outsourcing and managing outsource resources? And I'm like, it's the same, and I wanna hear your guys' thoughts on that, but like of the opinion of there's, the good thing about technology and learning is it's easy to do if you have an internet connection in a computer. So I still think there's a lot to, especially in American culture, like you need to translate business ideas into real world, you know, solve, or you need to solve those real world problems, which I think we are better positioned to do. But the coding aspect of it, anyone can do it. You just need to make sure that they're solving and doing the problem, solving it in a responsible way. Right. So I kind of leaned that way too, as you're starting to see before like, you know, maybe 10 years ago, you know, everybody wanted to try to do outsourcing, everybody goes and they do outsourced and then they realize that the amount of tech debt that they just inherited from those people were just insane. Yeah, and that's still see that today. Yeah. But it's definitely gotten way better. Like, so where you can have some of these development teams that are, you know, in, you know, India or Pakistan or wherever, and that they're actually really very capable people like the folks, you know, folks with one of my clients that pretty much only uses this development group out of Pakistan. And those guys are absolute rock stars. I'm blown away. I mean, even like the designs that they do, which usually historically hasn't been the case, right? Like you'd only let the new development, you wouldn't let them do design. Now I kind of come up with the high level kind of wire frame of what I'm looking for, and then I can give it to them, and then they go and they just nail it, and it's beautiful.
• What's the feedback loop look like with them, are they?
• So pretty much, it's pretty well locked down. So like if, you know, I can go into a Trello, and put in a backlog issue or whatever, and then I can just kind of watch it kind of flow through. I have access to all the Figma docs. So I basically have my own Figma. I do my designs, I give it to them, then they kind of bring it into their own, and then they start applying all their own stuff to it. And I just go in there and I kind of give feedback. And it's amazing.
• Yeah, and is this a small enough group that like you know all the--
• Yeah, yeah, yeah, yeah. So the other--
• And they're paid very well.
• Okay.
• So that's another big thing too, is like, if you're gonna work with them, you know, yeah, they're gonna tell you that they'll work for, you know, $30 an hour, but if you all of a sudden say, you know what, actually I'm willing to pay you 50 an hour, now you become way more important to them, right? Because they're not making that kind of money. And then all of a sudden they start making that money, they're gonna wanna love on you forever.
• Yeah.
• And so that's a good point. What I've seen not work in the past is leveraging outsourcing too much, you don't pay them enough. And then they just cycle out, the good guys somewhere else.
• Oh, totally. They'll stick the good. There's not that they're not bad developers, but they're gonna have one all-star. And Sean can speak to this much more, I think, but they'll have one all-star, they'll put front and center, they'll do the heavy lifting in the beginning and then they'll just, you know, just get away and you're like, "Oh, what happened to this good work?" And then the normal guys come in and just start butchering stuff. Yeah. Yeah, rolling over resources is tough. Getting people to do work for you and at a reasonable rate and having an ongoing long-term relationship is the best way to build software. It's much different than other types of projects because you build knowledge, that knowledge has to be shared. And you're much better off if you're, that's why people like to have a tech co-founder or somebody, or they want somebody that's gonna last and for us are almost all our clients are long-term. 'Cause they're like, well, these guys, we know we can rely on them, they know our stuff, they rebuilt it, they built it again, and they can keep going. Yeah, I think the other thing that I'm still, I'd love to get more, it just makes me wanna get out in the community and learn more, but you think that with AI and where we're at with our own productivity, where we've had a tested to our own acceleration of productivity, that you're seeing, I'm not necessarily like seeing people saying, you know what, man, I'm twice as fast as I used to be. And I think maybe it's the podcast, maybe it's just the excitement we have and all the stuff we've learned. But like, I wish developers were, I just don't think that it's popped into their head at how magical this brain that you have access to can be.
• So I have a couple friends and they're younger than me, right? So they're, I don't know, maybe, anywhere from 28 to 35 and they're developers who refuse to use AI. And it's just like, no, it's dog shit, not gonna use it, blah, blah, blah. And I'm just like, we know people age out of technology. Eventually they stop becoming a developer, they become a manager, they do whatever. And so I'm kind of wondering maybe that's these guys or like just this is their first kind of step into that aging out. But I couldn't believe it 'cause like a few of these dudes are like super brilliant developers and they're just like, I will not use AI. And it's just like that kind of seems very limiting.
• Is it because sometimes the code that's produced is kind of mid.
• Sure. Well, maybe. I mean, maybe, but like, but there's so many different things. It's not like we're just having AI write our code for us.
• I think that's the problem is people, and this is the same. We've all gone through this kind of life cycle of, "Oh, it's gonna do everything," and then you try to make it do everything, and then you're like, "Eh, actually might be better if I just did that myself." And then you cycle back, you try it again, and then you find where it fits into your workflow and actually adds value. 'Cause it's obvious. like it's novel at first, but when it's actually adding value, that's when you keep going back to the well and you keep using. And I'm finding myself doing it, but not as much as code dysfunction, but brainstorming or-- Right. Hell on the architect thing. Yeah, get me unstuck from errors kind of stuff. I will say so. I'm now a Claude. So I'm paying Claude now. And I finally got the artifacts thing figured out by getting to go into the settings and enable it or whatever. But then once I've done it, still, I don't like the overall user experience of Claude, I think needs a lot of help. Like, I mean, it's not bad looking. There's just a lot of clunkiness to it. Nevertheless, I went and basically said, hey, here's a WordPress plugin that I want because now I had to get back into a WordPress for another project that I'm doing. And I was like, I want you to create a WordPress plugin. Here's exactly what I wanted to do. And I actually just provided a user stories. So I just said, as a user, I want you to be able, I want to be able to see all the pages that were created off this template and blah, blah, blah, blah, blah, blah. And sure, now again, no non-developer would have been able to make this thing work, right? So that's a very important thing that you still have to kind of know what you're doing.
• I think that there's an opportunity for to formalize AI-enabled developer, right? And actually give specific use cases because I think there's brainstorming, architecture ideas, there's scaffolding out applications, which is nobody likes to do that, right? Nobody likes to create, okay, I'm gonna have to create 60 files today.
• That's why there's so many frameworks, right?
• Right, yeah. And I think that there's also, hey, outsourcing commodity stuff to AI, so your mental load is decreased. And I think that there's just a, I think that some people look at it and they go, well, they just don't know. They're like, I mean, you can't say, I don't use AI. You just can't. Unless you're saying, I don't like to be productive.
• Well, and I also think the advantage is like, we use it a lot to prep for the show. So we are forcing ourselves to use it and finding, so it's basically like, that's forcing us to find the use cases that work for us. If you're a busy developer, that's against a bunch of deadlines, you maybe don't have the like, freedom to play around. And if something's not making you more productive, you're gonna throw that tool away.
• That's very true. And I've been researching the, you know, some of the stuff that's going on with Microsoft co-pilot adoption, Apple intelligence. And I think one of the, if you're looking to get into this or you're looking to figure out who's gonna win, I've been waffling between a lot of the stuff. Have you heard it on the podcast? But I'm going back to chat GPT.
• Yeah.
• And I'm putting them way back into the driver's seat because I think an underestimated thing is actual understanding of user behavior. And they have so many users using AI. And they continue to iterate. And I just keep going back to that tool because of the experience. Yeah, the model might not be perfect. I know Claude is great and there's other ones that keep coming out. But based off of some of the experiences I've had lately, I feel like they're a lot further ahead. And tools like that, right? Tools like that. these model first companies that are coming out that are enabling these chat-based or other experiences versus which we've had in the other premise, which is, okay, all big tech's gonna win this. The personal AI race or the AI in the workforce race is gonna be won by Microsoft and Microsoft Co-pilot. It's gonna be won at the personal level by Apple and Apple Intelligence. And as time goes on and more people continue to use, companies like OpenAI, opening in 10 years might be a trillion dollar market cap company. They might not get acquired. They might be adjusted, I mean, the more I use that tool and the more other people do as well, the more stronger they're gonna get.
• Yeah. You know, it'll be interesting though, like what type of protections. So again, from my standpoint, again, I do like, I do like Foro as a model, right?
• Foro is every day.
• Yeah, so in Foro mini, which I just started playing with, which is literally like 80 times cheaper than Foro. I mean, it's crazy how much cheaper it actually is. And it's not meant to like, it's not so chatty. It's much more like a hike, what the hell is it? Claw at haiku or other small, like really small models.
• Yeah, but that one's actually really, really good. And I think there's a ton that you could do with that. Like you could multi-shot like 15 different queries to it and just like do all this crazy stuff and your costs are gonna be really manageable.
• Well, yeah, it's actually cost effective now to personalize ads per individual.
• Yeah.
• So like Google has--
• That's a great idea.
• Well, I read it somewhere. (laughing) But the point was like, well, where it's like, all right, Google knows my shopping habits, viewing habits, and they can throw that into a model and they have their own mini models, but it's cheap.
• It's like when Tom Cruise walks into the retail store in Minority Report and he'd still on that Asian guy's eyes. Do you remember that? And it's like, hello, Dr. Quasi-Mato, would you like to wear those cocky pants again?
• Yeah, no, we've basically created Minority Report. The only thing I'd really, you guys use Google's a personalized search result thing. Okay, so I love it because it's so relevant 'cause it knows exactly what I've been looking for. What I don't like is I accidentally leave it and come back and it refreshes the whole page. And there was something at the top that I was like, "Oh, oh that looked interesting." And I was like, "What am I gonna do? Go look for it." (laughing)
• That's the worst, honestly. Like, and that's one thing I hate about these intelligent feeds.
• It's that same exact problem, right? Like then you accidentally refresh or you do something and now it's gone. You're never gonna see that piece again. It was just like, no.
• So, some news, okay, real quick, shifting gears. We had kind of this like, oh, big news week, but also come with a dud, okay? We're gonna start though with Apple Intelligence. All right, so iOS 18.1 is out, it's a beta. Also Mac OS, Sequoia beta four is out. So if you wanna try Xcode 16 with and actually build apps with their new code completion tooling in Xcode, you can do that. And obviously build, if you wanna build anything to the iOS 18.1 anyways, you have to upgrade that anyway.
• Oh sure.
• The actual Swift SDKs are not available. So one of the big things about Apple Intelligence is it's gonna come with these SDKs that allow you to leverage the image playground, it allows you to edit, yet utilize the writing tools inside of your apps. Apple's play there is get these APIs in the hands of developers, have developers enhance their apps, kill Grammarly, and enhance the entire ecosystem. That's generally what they do. The actual operating system though, 18.1 was the first time where they introduced Apple Intelligence and you have to get on a wait list once you download it. But I did that and I got off the waitlist and actually almost instantly. And I mean, I just couldn't believe how limited it was. I think that there's an underlying base layer that they're putting in here when it comes to the on-device AI, the models. But Siri was just as dumb. Well didn't you say it out put a result with Markdown? (laughing) And didn't, like it just didn't even format it for its own reader.
• So the only writing tool app that has been updated is the Notes app. Okay, and what you can do is you can select your content. You have a new, you know, it's all integrated into Apple's core UI features, right? That's the great thing about Apple is they're gonna build this in, there's not some new button or anything. It's built into the existing way you use apps. It's not gonna be some, necessarily some magical new thing. They do have a little icon that they're using for AI similar to what Microsoft is doing. Those little-
• The stars.
• The stars for it.
• It's a different icon, but-
• Right.
• But any app that has text is gonna have this writing tools overlay, and the writing tools overlay has different AI-enabled tools like summarize, make more professional, make more loose, whatever. Well, if you select a group of text, it's doing that on device AI and summarizing it, but it was returning the raw markdown. So you were giving it formatted text and they released this in a beta.
• Yeah.
• I hope that this forces them to then make the notes app basically just be able to render out Markdown and let me even write in Markdown kind of like what Notion does, right? You can go, you know, and you can even enable that in Google Docs where you can write it.
• Which doesn't work all the time.
• It doesn't.
• Especially in copy and paste in Google Docs.
• In Google Docs.
• Notionalies.
• Notion, dude. So here's a little tip for you guys. 'Cause one thing that sucks about using any of these large language models is they all output markdown, right? Like 99% of their output is all markdown. So when you copy that and then you go paste it into an email, you get all your markdown format and you're like, well, this doesn't look right. So what you do is you copy it, you go paste it into notion and then you copy from notion to wherever your destination is and it's perfectly formatted.
• Go ahead Mark down to text tab open on my computer last month, so thanks for that tip.
• Yeah, VS Code has a good plugin as well for that. But also what I've been doing, because I have an automation that's running every day for my calendar, I have it, I just say like format those for email instead.
• Oh, that's a good idea.
• But that's extra typing, like, yeah.
• Well, so side note, let's, because we're fucking having a purpose today other than chaos. Tell us a little bit about your calendar automation that you set up.
• Oh yeah, so I have a really bad memory and I wanted a daily reminder of events for me and my girlfriend that are coming up for the next couple of days. So I'm using an open source tool called Active Pieces.
• Yeah, which is awesome. Really, it's like make.com, but it's open source. You can run it on.
• Active pieces.com.
• Active pieces, yeah.
• Is that just the name of the project?
• That's the name of the product.
• Yeah, just search for it. They have a, so they do have a cloud paid version, just like Zapier or make, but they have a self-hosted version, which I'm hosting on a computer just running at my house right now, which I love. But anyway, so I have this automation that it, on a schedule every morning at 8 a.m., it goes and gets my next 10 calendar events, which I guess I should probably adjust that to be next seven days. But anyway, it does that, and then it makes a list for me, but it uses chat GPT-40-mini.
• Oh, yeah, okay.
• To get it like a one word description of like what's coming up.
• Wait a minute, wait a minute, wait a minute. Okay, so you're only using the, you're only using chat GPT to generate your g--
• No, the first thing I do--
• Okay, all right, all right.
• So I have it like it's like seven steps. So first thing I do is like give me a one word description of what's coming up and I send that to Giffy to or whatever to get me a GIF or GIF for the top of the email. That's like the most fun part is like me and my girlfriend like, "Oh, what's the give of the day?" And then I send it to chat GBT to format some things. And then I have like some JavaScript that's like just taking that format of data and actually outputting a tissue.
• But isn't that a great example of, but I don't think this is another thing people might not realize use cases with AI is a large language model is great for inserting itself into workflows.
• Oh yeah.
• Right, because you can take each, on a workflow you have an output from each step.
• Yeah.
• Right? And you don't want to go code the man map fields and interpolate the responses in code, you just wanted it to figure it out. You can tell the prompt, you can say, okay, output one, send it, send, it can kind of reason a little bit about what needs to get sent to the next step or it can summarize it or it can make it pretty or you know what I mean? The first step, that eliminates a lot of problems. Yeah, the first step I did just say, take this structured data and write it into an email. Then I thought, well, instead of using chat GPD every time to do that, I just had it write the code that took, because it comes back in JSON from Google calendars or whatever. So I'm still using it and it's great. Like, yeah, I'm pretty proud of it.
• So, just other takeaways from Apple Intelligence. It is highly integrated with Siri. Siri now has this amazing beautiful animation around the outside, which you can see.
• No, it's just cool.
• You sure the camera.
• So you can see that as we're talking to Siri.
• Stop.
• You know, there's very limited feature set, which is a little concerning considering this is a beta for or whatever it is. There's a, there's something that they're calling AI that was very interesting to me. I could not get it. I don't believe it was available on iPadOS 18.1, but you know how you have focus mode. So focus mode, you kind of pick the apps, in your different modes that you might want to get notified by.
• Right.
• 18.1 ships with a feature that will basically say, hey, go AI focus mode. And it'll only escalate push notifications that it thinks are important. Which might be good for you if you still have Gmail notifications turned on. (laughing) And maybe I would consider doing that because I do want to get notified sometimes of a Slack message or a Gmail. And for years I turned that off and I turned it back on 'cause I was like, you know, this is making me less efficient.
• Yeah.
• Yeah, I turned them all off and like, of course I do miss emails from time to time, but it'd be nice if something bubbled up that was like maybe more time-sourced-
• Yeah, like fraud alert, you know?
• Yeah.
• So it's, that is the, we're gonna get to that. That's the missing piece that you're just not seeing, and you're gonna see that when developers, 'cause I think they're just still putting the foundation in for the house, for the hotel, for the skyscraper, the software's not caught up yet. I think what's really important to note is one, it's on device. So developers, there's not an increase in cost to the software, maybe a little bit of a battery usage cost, but they're laying the foundation for, like, to enable all of these apps to be, use some pretty powerful things. Maybe not this generation, but maybe the next. And they're also making everybody upgrade if you want.
• Well, you have to fit 15 max or go.
• So literally, I've, I'm--
• Yeah, I mean, that's the reason why I got this iPad. I'm a fan boy.
• Yeah, mine's open.
• So forcing, here's the issue with Apple. Forcing people into an upgrade cycle to get a suite of tools that isn't ready for mass user adoption could really hurt their brand. Because I do think that Google shipping their mini model, their ability to, you know, they've built a little bit more around the AI tool chain and space than Apple has. And I think that, I think that, I think it, Apple's gonna really rely on their developer ecosystem to build this stuff out. And I don't think it's gonna be ready in a couple months. I think it's gonna be--
• It's gonna be wild.
• I think it's gonna take a while. I think that Apple's slow approach may end up winning, but in the short term, (laughs) I think that's gonna, I think they're gonna take some lumps.
• Well, so like think about the, so Chad GPT, the voice thing that we were talking about, about three months ago or two months ago, whatever it was, where we're all excited about the legacy period. So excited. Yeah, and but they, they were like, so they always blamed it on safety. Now the reason we couldn't release it right now is safety. Their blog posts are so elitist on their website. Oh yeah, they're always like, we had to refine it so it was ready for, you know, really, really ready for our users. It was not polished enough. And just the way they talk, you know, like that thing they posted with Elon and all that stuff. Oh yeah. But yeah, chat GPT voice, they said, oh, it's out.
• Yeah. - News comes out.
• No. - Click bait, click bait, click bait.
• No, it's not out. - And then it's not out.
• They're rolling it out to people that, you know, the 1% of their user base.
• Yeah, but I think what happens is these companies get so confident. And then they're just not thinking that, again, we're putting like these little thinking programs who can kind of act on their own into our systems, and you don't have necessarily the infrastructure, the QA processes or anything for that to really be able to put it through its paces. and then some QA guy just types in something and goes, "Uh-oh, boom." It's like, "Oh my God, guys, hold on a minute. We can't do this." And so it scares me that one, like, you know, companies like Apple or even Google or Samsung and the stuff that they're trying to do and they're saying, "Yeah, this is gonna be coming." That, "Yeah, we're gonna, we're just, we're kind of back in this wild, wild West time." And I kind of like it.
• Yeah.
• Well, I agree. I think that there's, we're in this weird time where, yeah, the kids are going back to school, everybody's announcing all this stuff, but it's not really there yet. But it's all stuff. I'm just based off chat, GPT voices, the existing functionality. I think that it's a fantastic tool.
• Oh, to 100%.
• I mean, you could, I could see myself having that her type of conversation.
• Yeah, I do it when I'm driving. Like if I have a new finger, I kick it off and I just start having a conversation and just kind of work through. But again, the latency is kind of annoying. But like if you wanna have a real conversation, so that's why like I was so pumped for that latency zero and it was just like, just give it to me. So every day I go in, I check and I still don't have it, but yeah.
• Changing the subject a little bit, but I forgot, I played around this this week. Have you guys cloned your voices with any tools yet?
• No, no, Jacob did it was damn scary.
• Was it? - Yes.
• 11 Labs, which by the way, real pro tip for a real world usage, Sean was building an app that he wanted a sound effect for, and 11 Labs has an amazing sound effect creation tool. So it's like you want something very specific, it was like a poppy sound.
• So we were building a barcode scanner, we built with chat GPT, and Rust, and I don't know anything about Rust, and it works perfectly. So people don't use that.
• Not only Rust, though, it was Rust too.
• It was Rust too. - Rust was to Rust.
• To Rust, and then Rust to Wasm.
• Rust to Wasm, so you can run it in the browser.
• I don't know anything about those technologies, and I just shipped a fully working barcode scanner that I could probably sell and license.
• Yeah. And the sound effect came from 11 laughs.
• Yes.
• That was my contribution.
• It was perfect though.
• Yeah.
• That's perfect.
• Yeah, that's very cool though. So how long did it take for it to train your voice?
• Oh, just one 30 second sound bit.
• Yeah, it sounded a lot like him.
• Yeah, I would say like I was talking like very, I was just reading something on the screen and so I was talking very--
• You didn't have as much of your personality.
• Yeah, if sound, I mean, I would not be able to tell say what's wrong with it.
• I wanna play around with it and having a little bit more like energetic conversation 'cause I was just reading stuff off the screen and it very much sounds like me being like, "Hello, I am very happy."
• Yeah, but--
• Well, can you just take some snippets from the podcast?
• Yeah, I could, yeah, I could. And I was gonna, I'm not gonna do it 'cause that's mean to my mom, but I was gonna like write up a script and trick my mom.
• I'm in the prison in Mexico.
• Send me $30,000.
• My mom would run to Mexico. (laughing)
• Oh, that is so bad.
• Speaking of which, so if you go to on iOS 18.1, if you do this, it pops open, a chat GPT style conversation with Siri.
• Whoa.
• Yep, so there's two ways to use Siri now, and this is, I think this is where they're heading, where you pull up Siri and you have that chat style
• Siri, yeah. - conversation with her. And it doesn't work well. Like how are the alerts going? You kick me off everybody's Siri by the way. Yeah. (laughs) Yeah, it can help me find files faster. Oh my gosh. Oh, that's, yeah. I mean, it-- That's cool. It can just quickly quickly start a chat conversation with Siri. Right. But all of the responses, and I don't know if this was a good idea, you know, they're product, they're so smart product-wise. All of their responses are not, it's not her talking right now. It's them giving you back the data in that way that it wants to do that.
• Right.
• I don't hate that. I think that the chat responses is very rarely the best way. Like, you know what I mean? Like when I'm talking to my Google Home Assistant at home or oh my God, at my house, it's like home. I'm trying to get a recipe or something like that. And it's just like, according to recipes.com on February 20, I'm like, okay, just tell me the ounces or whatever the hell I ask, you know. It'd be nice to have a screen associated with, now that's obviously a limitation of an audio or a speaker.
• But there's no doubt that, like, well, so you had the Amazon home, they came out with the little screen one, I think. Google has one, surely Apple will have the Apple home, or the HomePod, I guess is what's called?
• Yeah.
• I love, yeah, I love my HomePod. So are you pro, do you like that it's more, or more like visual, or do you want it to be serious talking about you?
• So I didn't trust my own opinion on this Apple intelligence review. So I went and watched a bunch of other ones on YouTube, and there were some people that were really seeing it, really seeing the potential of it. And like, I can really see this integrating into my workflow. Now a lot of them were using the phone, which is, you know, your phone is like your live and die. And I think that some of the features are more built out on the phone than an iPad experience. But I just of always loathe Siri.
• Yeah, Siri's always been really tech.
• It's so bad.
• Yeah.
• And I've used chat, Sheepy-T voice, and I like it. So what's the dang difference?
• Yeah.
• So anyways, can click. I think it's way too early to tell on Apple Intelligence. They're advertising all this stuff that doesn't exist, app integrations, you know, tellin' Siri--
• Jetology.
• Yeah, Jetology.
• Yeah, that's not in there.
• Image playground's not in there.
• So no way to even get a judgment team.
• The only things that are in Apple Intelligence right now.
• On the iPad.
• On the iPhone, iPads even less. On the iPhone you have, Summarize a webpage with Safari. You have Summarize an email in mail, Priority inbox, Writing Tools and Notes. You have an enhanced Siri. And then you just have the overall opt-in to in Apple Intelligence workflow and like, you know, on, you know, whatever it did to ship the model.
• Does the Safari here then have the, on the iPad? It does, it does, not on the iPad?
• Nope.
• We were talking about this earlier this week, but do you think there will ever be a day where they're just listening to us all day long? And then at the end of the day, they can be like, hey, Jacob, you've mentioned you needed to do X, Y, and Z.
• Absolutely dude, that would be a killer feature too, especially if it's just on device, right? That's the offer.
• Yeah. It would scare me, I mean, it scares me either way, but it would scare me a lot more if it wasn't on the run.
• Yeah, no, it has to be on-device.
• Secure on-claves, if it even ships, I think is enough to lay the foundation for an unbelievable amount of things that you would let AI do that you otherwise wouldn't. 'Cause it's secure and it's on-device. And it's completely obfuscated. And you know, you're working with privacy, how important that is. We were talking about it yesterday. How do we operationalize AI and a SaaS product where it's like HIPAA times five. Yeah. You know what I mean? And I don't think people have a good answer for that. This is a good segue. Okay. Right? This is a good segue into the LAMA 3.5, oh yeah. 450 billion per minute or model that they just released. Because I think that that is going to be what people ultimately build and run in their own places. Because again, you can't run, You can run chat GBT in your Azure thing, or you can run it wherever. But yeah, with Facebook just released this model called, I guess it's 3.5 or Lama 3.5, it's 450 billion parameter model, that according to the benchmarks, it has it competing with chat GBT4.0 and Claude 3.5 saw it, right? Which are two of like the latest things. Now again, this thing is massive. You're not just gonna be running this on your little pool.
• In Zuckerberg just wants to give it out. They also claim that it's more energy efficient by the way. Really? Which I thought was interesting. Which gets back to our main premise, never, ever, ever, ever underestimate a Facebook engineer. No, no, seriously. Never. Yeah, and so now you can interact with it if you go to meta.ai if you're a meta-logger. But you can use it there. You can use it on Grock. And right now it's still a little slow because they probably have to optimize it for it or whatever. that thing running and be really fast, I think that that could be an absolute kind of game changer. So who's our top three model? Well, going back to the premise that model makers are learning faster, are doing better, are more primed, which we've also had the other premise, which is models or a commodity. Go back to the other premise, which is model makers are the leaders. Who's your top three? Who are your top three companies? Well, if you just look at the chart, I think it's meta's number one, four O's number two and ultra's number three.
• Meta is over four O.
• Are they?
• I think that, well, it was their numbers, their benchmarks, but yeah.
• Everybody always releases these benchmarks, so they always up there.
• Yeah, we've got OpenAI, we've got Anthropic, we've got meta and we have mixed Mistral.
• And so Google's not even the top five.
• Oh no, but ultra is right up there, but yeah, I would--
• Did you use it?
• Yeah, it's pretty awesome.
• So Meta is publicly traded, San Francisco, open AI nonprofit San Francisco, kind of profit, and profit, non-profit. Private, funded San Francisco is Mr. Prince.
• Mr. Prince.
• Yeah, their French.
• That one was open source, but only for personal.
• Yeah, and then they just released a new large model. It's only for research. So, Mr.ill kind of did a rug pull, like Mr.ill kind of came out and said, "Oh, we're gonna be the antithesis of like open AI and we're gonna make these open source models.
• Yeah.
• Exactly, like almost exactly the case that open AI started. We're gonna be an open source, we're gonna make this, and then they start seeing some of it, and they're like, "Ooh, maybe we shouldn't make this open source." Then they probably see the dollar signs, or like, "We definitely shouldn't make this open source."
• Well, there's such a, that's where you go to the other side of it, which is where's the money. Look at perplexity, right? They were trying to do this revenue sharing thing, and all that, and like, are they?
• I'm still kind of curious, 'cause there was an announcement recently that Perplexity's partnered with a handful of other news outlets.
• The Atlantic, yeah.
• You talked about that.
• Her way was the Atlantic OpenAI.
• The Atlantic was open AI.
• You're right.
• Yeah, maybe it was like the New York Times, I don't know, maybe I would have heard that one. But the point was, I wanna talk more about the search GPT and how that's basically a rip off of Perplexity later, but like, but Perplexity--
• Or even what Find was doing from a coding perspective.
• Yeah, yeah, Find was great. Do they still exist?
• Fine.
• Oh, come on, give us a drop.
• They still exist, right?
• Do they get bought?
• I don't know.
• I love fine.
• Yeah.
• PHIND.
• It's great for coding.
• Perplexed the, what is a search engine, an AI powered search engine, basically, where you, you know, they go and get, look up real web pages in real time and then try to distill that information and their results. Now they got in a little bit of a heat because they were basically like, there's no attribution. And then they started doing attribution. Now they've got revenue sharing they're trying to do. which is nice, I guess, but is it ever gonna be enough?
• Well, I think that there's a notion that if the search GPTs of the world, they know that they can't train on artificially created data forever. That the web has to continue to exist and content has to continue to get created. And so they need to not just, in my opinion, just harness all the data and just own all the data and don't credit anyone for the data and don't compensate anyone for the data 'cause eventually they can run out of data.
• Yeah. I think there's a whole dark, dark data space out there that like, so one of my clients was talking to a this big consulting company who's wanting to build like a lot of AI things for their specific industry. And they're like, we're willing to partner with you guys. If you'll give us some subject matter experts and you'd give us like a bunch of data to help train your model on, and the client's like, "Well, let's hear about this." And kind of, and I'm just like, Oh, like you're giving away your most valuable stuff. But I think that we're going to start seeing that as these other companies kind of start running out of just public web training data, that they're going to start doing deals. They'll start doing deals with companies that have this data that's hidden behind their fire. Well, and that data has never really been invoked into the public space. No, but if they came and said, hey, listen, a company, we'll pay you $30 million if you you let us train on your data.
• Yeah.
• How many American companies would say no to that?
• Right, well, Salesforce's whole ad's going on right now at the Olympics as a bad guy AI. So they're positioning themselves as they will never ever sell your data in the exact situation you're talking about.
• And it's on a board at a corporation right now. Number one concern is privacy and security. It's not adopting AI. because they know that if shit hits the fan, it's on them. And they know that there's this leak concept and people are getting, I mean, crowd strike, we've got all these different problems. It's, the risk is real. And that's why you're seeing actually the anti-AI space is coming into play. And that's why for me, it's right now it's big tech trying to put their tools into enterprises, right? 'Cause you already do business with them. And it's that mid-market SaaS companies, like the ones you've been talking about and looking at where a guy that has minimal risk or maximum risk threshold, he's willing to take that just to be more productive to go hunt down a million leads on LinkedIn or write better emails or generate better images. You know what I mean? But from a, what's up?
• I wanna comment on this, you guys can cut it out.
• No, please.
• I wanna hear your guess of state.
• Yeah.
• That's a commercial, the Olympic commercial of Google asking. So he was like, my daughter really looks up to this Olympian who write a letter helping her write a letter to the Olympian and everyone's freaked out.
• Yeah.
• She's actually getting a ton of backlash.
• Yeah.
• So the question from our producer was there was a commercial on the Olympics about a, I did see that and that was, that did, it was a real world personal AI use case. which is someone adored this athlete and they wanted to write a letter. Instead of writing it themselves, they just asked AI to do it. And sure enough, it did it. In fact, this morning, a kid asked me for a letter recommendation for college. Chat GPT wrote it. I gave it some pointers. But that was a very simple use case of somebody.
• They're very upset.
• What were they upset about?
• Well, God.
• You think negative about AI, that adds kind of showing that. like taking out the personality, not letting the daughter who's critical thinking about some objection was up to, but letting the computer explain the truth.
• I think there's very valid.
• That is a valid argument.
• I heard people make that argument and my knee jerk was like, shut up. Like, it's helping someone who's not a great strong writer communicate to someone that they look up to. But I actually, after I thought about it for a used critical thinking, and I was like, okay, they have a good point because there is something very authentic about a child who can't write that well, and they're sitting there struggling like, "Mommy, I wanna say they're great." And I just actually saw a letter to my grandma when I was in third grade or something. Horrible writer. I was horrible, but it was very cute. 'Cause I was very--
• Well, the letter I wrote for example this morning, that letter of recommendation, it was really cheesy. I mean, I probably could have worked on it a little better, but it conveyed the point it was well-structured, but it wasn't gonna pull any tears.
• Yeah.
• Yeah.
• And the concept of the youth utilizing generative AI, especially for writing is now. Now, I think about it as a concern to me because I became a good writer by having to write this brute force.
• Yeah, and it doesn't take away the value where it can help you. Like, I'm trying to write a case study about a project that we worked on. So I give it a bunch of details and I say, give me an outline of what I should do. I pick and choose. it's like you should write about how it, the overview of how it affected the product or the client. That's where I like it, 'cause then I can put in my personal, like the real world or real life experiences into that. And I don't have to sit there and look at a blank screen for two hours and think, oh my gosh, I suck at writing. But yeah, I think it goes too far.
• But let's say that I start typing and I'm writing my letter and I just start hitting the next word, next word that the AI or that the type ahead is giving me, right? How is that any different?
• Yeah.
• Like, I mean, you can make an argument that it's different. So nevertheless, I do think that the children thing does scare me a little bit because the fact is, is that we learn to write and to communicate by writing.
• Well, we've never, I mean, we've never lived in a world where you didn't have to write.
• Yeah, but let's take Rust, right?
• So right now, people would just go plagiarize off Google.
• Yeah.
• And I had people doing that in college in grad school that had never written 'cause they were from, they weren't not from this country and all they did was math and science.
• So, okay, so to that, the rust, like the rust work, right? All of a sudden there's some bog in the decoder, right? How are you, without AI,
• Screw it, you just send it back to AI.
• Yeah, you just send it back to AI, but nevertheless, you have no mental model of that code, right? Like if we write it from scratch, we have that mental model.
• Yeah, well, and it's the same thing. Like I was using a lot of the no code solutions this week. And I thought, my God, AI is gonna be so good in a couple of years at, I wanna type it, type out, hey, I wanna workflow that does X, Y, and Z. And because it's using all these little modules, it's gonna know exactly what code to plop in there, right? I have no freak, it doesn't matter.
• Exactly.
• Do I need to know, do I need to know the ones and zeros? Like I studied computer engineering. and computer architecture is insanely complicated. No one needs to know that anymore, except for a select view, right?
• There's only one word that comes into mind for me with all this, and it's productivity. I don't give a shit what you say. I may be an unauthentic person. Yeah, you should have written the letter, girl. That was your hero. You should have, you know?
• You were so good at telling us this girl in this commercial.
• She should have written the letter.
• It wasn't, it wasn't, it wasn't, it was like the dad.
• There's the damage.
• You lazy son of a bitch. (laughing)
• But you know what? You know when somebody asks you an uncomfortable question in an email or you're trying to get, you got to get home and cook dinner. And it's 4.44 and you need to send a couple emails. You know what, it's about productivity. Pick and choose your spots, but the whole concept of going, I'm not gonna use this stuff because I think it's, I just don't like it, I think blah, blah, blah. I think you're missing out. It's also on life. It's a somewhat like, I don't know if it's elitist or I get the feeling of being like that is unauthentic. The dad should have helped the kid. It's still a moment in their life that they could have looked back on and had this wonderful experience. But just because that wasn't a great application of it doesn't mean we should just throw it out.
• Yeah, no.
• Like no use for it.
• I found like, and this was maybe, I think early last year, but there's a thing where I had a really hard time trying to communicate with my daughter on this topic. And every year we go through this and I'm just like, "Oh, this is gonna suck." And so anyway, I go to chat GPT and I kind of just lay out the scenario because anything that's like that you have emotions involved with, right? And you just can't necessarily think very clearly. I find it was pretty, and it broke it down. It was just like, she's growing, she's gonna be independent, she's gonna find herself. She wants to separate herself because she needs to go find blah, blah, blah, blah, and broke it down to the way. I'm like, oh my God. And I'm like, well, how can I convey that I understand this now?
• Yeah.
• And he, to hear, just say this, you know? And it was, and it was just like, we had a moment, like for the first time in 20 years that I was able to step back and get out of my own ego because chat GBT helped me kind of see it from her perspective. And so for that was just absolutely invaluable.
• Yeah. that there is a lot more difficulty in accepting AI that I'm seeing from normal people that I would not have expected. 'Cause when productivity tools come around, people usually adopt them. And to me, AI is a productivity tool. It's not a person-replacer. It's not a heart-replacer.
• Right, right.
• And I think it's teach his own, but I'm gonna be out here getting shit done.
• Exactly. So yeah, I can't take a quick rest. I think we're. We're going to wrap it up. We're going to hold it. I'm just going to pee my pants in. I should have taken a mood. Jana cheese podcast. We'll see you next week. Later. Thanks.